A Novel Approach to Extensive Physically Unclonable Functions Characterisation Leading to Secure Challengeless Authentication

With the rapid proliferation of Internet-of-Things (IoT) applications, there is a growing need for authentication mechanisms that preserve prover-side simplicity while avoiding exposure of challenge-response structure. Physically Unclon able Functions (PUFs) have attracted significant interest due to their intrinsic randomness and hardware-level uniqueness, which are commonly exploited through Challenge-Response Pair (CRP) protocols. However, such conventional approaches expose the challenge-response structure to adversaries and typically require error-correction mechanisms to compensate for noisy responses, adding both design overhead and additional security considerations. In this work, we propose a novel challengeless authentication framework that eliminates the need for explicit challenge transmission during authentication and avoids CRP based verification. We introduce the concept of Topology Propagation Patterns, which capture latent PUF-specific features through structured response sequences generated internally by the PUF device. Authentication is then achieved using a verifier-side convolutional neural network model with 6.13 million parameters. Unlike helper-data-based reliability schemes, the proposed method does not rely on error-correction-assisted reconstruction during authentication; instead, it uses response-derived images directly as the basis for classification. Our approach requires no modification to the PUF structure itself; instead, it alters the logic used to generate challenge sequences from prior responses, thereby preserving the integrity of the original hardware implementation. Experimental results on Arbiter PUFs, 2-XOR PUFs, and 3-XOR PUFs implemented on Zynq-7000 FPGAs demonstrate high classification accuracy, reaching nearly 100% in 5-device scenarios and over 95% in noisy operating conditions for favorable settings, while also revealing the practical trade offs among PUF complexity, device-group size, image size, and authentication bandwidth. Under the stated threat model, the proposed protocol is secure against conventional CRP-based modelling attacks because explicit challenge exposure is removed during authentication and the data relationship required to construct labelled CRPs is no longer available to the adversary.